Design Of Garbage Classification And Guidance Device In Public Places

It is of great significance to reduce environmental pollution,save land resources and improve the quality of life to carry out the garbage classification system and accelerate the establishment of the garbage disposal system of classified delivery,classified collection,classified transportation and classified treatment.The classification and collection of garbage is a key link in the garbage treatment system,which can effectively achieve the maximum extent of waste recycling and reuse,and lay the foundation for the application of advanced garbage treatment methods in the future.The use of garbage classification guidance device for public places is helpful to reduce labor costs and improve efficiency in the process of garbage classification and delivery.In recent years,with the development boom of artificial intelligence technology in the world again,deep learning technology has brought great breakthroughs to the fields of image classification and object detection.At the same time,the performance of embedded hardware devices has been greatly improved.Based on this background,a garbage classification and guidance device in public places is designed based on embedded hardware platform.In this paper,42 types of garbage that need to be detected are determined by analyzing the common types of garbage in public places,and a garbage image data set is constructed by filtering,sorting,labeling and expanding the garbage images obtained from the datasets of the current authoritative garbage classification contests.Then the garbage detection algorithm was studied and trained,mainly including: selecting YOLOv4 as the core algorithm of the device through the research and analysis of object detection algorithms,developing the YOLOv4 garbage detection model by using YOLOv4 object detection algorithm and garbage image data.Finally,the hardware and software design of garbage classification and guidance device is carried out.Firstly,the hardware system was designed and built,including NVIDIA’s Jetson Nano hardware platform,touch screen,camera,camera stand and other auxiliary devices,and then the system was built.Then the garbage classification and guidence program is designed and implemented with YOLOv4 garbage detection algorithm as the core,including the design of the function and process of the application program,and using Py Qt5 to design the graphical user interface and integrate each function module.For the core garbage detection module of the program,Tensor RT is used to accelerate the reasoning process of the algorithm in Jetson Nano platform.After experimental evaluation,the mean accuracy(m AP)of the YOLOv4 garbage image detection algorithm in the test set reached 90%.For 416*416network size,the forward reasoning process of YOLOv4 algorithm is about 2.865 FPS after Tensor RT acceleration,which is improved by about 44.6%.For the input image with 416*416 pixels,the designed garbage classification booter can display the recognition result at a speed between 2.52 FPS and 2.75 FPS from the input image to the display.The device is designed to meet the design requirements of precision and speed,and has enough stability and good user experience.